Wen Tao Wu expanded zero-shot dataset support in the embeddings-benchmark/mteb repository by adding PatchCamelyon label files, enabling more robust evaluation of zero-shot image classification tasks. He introduced a new labels file containing 'lymph node' and 'lymph node containing metastatic tumor tissue,' improving the benchmark’s coverage and reproducibility. The work involved careful dataset curation and integration with the existing Python-based benchmarking suite, ensuring consistency across evaluation runs. By focusing on dataset structure and compatibility, Wen Tao Wu addressed a gap in the benchmark’s zero-shot capabilities, delivering a targeted feature that aligns with broader goals for model evaluation and reproducibility.